Examples of probability mass functions (discrete variables)

Introduction to theory of probability and statistics

Lecture 6.

Examples of probability mass functions (discrete variables)

prof. dr hab.inż. Katarzyna Zakrzewska Katedra Elektroniki, AGH

e-mail: zak@agh.edu.pl

Outline ^:

●

Definitions of mean and variance for discrete variables

●

Discrete uniform distribution

●

Binomial (Bernoulli) distribution

●

Geometric distribution

●

Poisson distribution

MEAN AND VARIANCE OF A

DISCRETE RANDOM VARIABLE

Mean and variance are two measures that do not uniquely identify a probability distribution. Below you can find two different distributions that have the same mean and variance.

MEAN AND VARIANCE OF A

DISCRETE RANDOM VARIABLE

The variance of a random variable X can be considered to be the expected value of a specific function of X:





 X 

X

h ( ) (

In general, the expected value of any function h(X) of a discrete random variable is defined in a similar manner.

Discrete uniform distribution

The simplest discrete random variable is one that assumes only a finite number of possible values, each with equal probability.

A random variable X that assumes each of the values x₁, x₂, …, x_n with equal probability 1/n, is frequently of interest.

Discrete uniform distribution

Suppose the range of the discrete random variable X is the consecutive integers: a, a+1, a+2,….b for a≤b.

The range of X contains b-a+1 values each with probability 1/(b-a+1).

According to definition mean value equals to:

Two-point distribution (zero-one), e.g. coin toss, head = failure x=0, tail = success x=1, p – probability of success, its distribution:

x_i 0 1

p_i ^1-p p Binomial (Bernoulli)

where 0<p<1; X={0, 1, 2, … k} k – number of successes when n-times sampled with replacement

For k=1 two-point distribution

Examples of probability

distributions – discrete variables

n k

p k p

p

n  

( 1  )

,  0 , 1 ,  ,



 



 

Examples of probability

distributions – discrete variables

Binomial distribution

Binomial distribution - assumptions

Random experiment consists of n Bernoulli trials :

1. Each trial is independent of others.

2. Each trial can have only two results: „success” and

„failure” (binary!).

3. Probability of success p is constant.

Probability p_k of an event that random variable X will be equal to the number of k-successes at n trials.

n k

p k p

p

n  

( 1  )

,  0 , 1 ,  ,



 



 

Pascal’s triangle

2 1 2 2

1 1 2

0 2 2

1 1 1 1

0 1 1

0 1 0 0

 



 



 

 

 



 

 

 



 

 



 



 

 

 



 

 



 



 

n n n

!

! ) (

! k k n

n k

n

 

 

 

 Symbol 

k n n k

k

n

a b

k b n

a







 



 

 

0

) (

Newton’s binomial

1

1 1

1 2 1

1 3 3 1

1 4 6 4 1

1 5 10 10 5 1

1 6 15 20 15 6 1

n = 0 n = 1 n = 2 n = 3 n = 4 n = 5 n = 6

+

Pascal’s triangle

Bernoulli distribution

Example 6.1

Probability that in a company the daily use of water will not exceed a certain level is p=3/4. We monitor a use of water for 6 days.

Calculate a probability the daily use of water will not exceed the set-up limit in 0, 1, 2, …, 6 consecutive days,

respectively.

Data:

6 , ,

1 , 0 4 6

1 4

3    

 q N k

p

0 6

1 5

2 4

3 3

4 2

5 1

6 0

1 6 3

) 6 (

6

4 1 4

3 5

) 6 5 (

5

4 1 4

3 4

) 6 4 (

4

4 1 4

3 3

) 6 3 (

3

4 1 4

3 2

) 6 2 (

2

4 1 4

3 1

) 6 1 (

1

4 1 4

3 0

) 6 0 (

0





 

 

 













 



 



 



 



 



 







 



 



 



 



 



 







 



 



 



 



 



 







 



 



 



 



 



 







 



 



 



 



 



 







 



 



 



 



 



 





k P k

k P k

k P k

k P k

k P k

k P k

k P k

Bernoulli distribution

9 9 9 1

3

356 . 0 ) 0 ( 4 1458

3 9 9 6 4

1 4

6 3 ) 5 ( 5

297 . 0 ) 0 ( 4 1215

9 9 15 4

1 4

15 3 )

4 ( 4

132 . 0 ) 0 ( 4 540

3 9 20 4

1 4

20 3 )

3 ( 3

033 . 0 ) 0 ( 4 135

9 15 4

1 4

15 3 )

2 ( 2

004 . 0 ) 0 ( 4 18

3 6 4

1 4 6 3 ) 1 ( 1

00024 .

4 0 1 1 1 ) 0 ( 0

6

6 1

5

6 2

4

6 3

3

6 4

2

6 5

6





 

 

 











 



 

 



 













 

 

 



 













 

 

 



 













 



 



 













 





















P P

k

P P

k

P P

k

P P

k

P P

k

P k

Bernoulli distribution

0,00024 0,004

0,033

0,132

0,297

0,356

0,178

0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4

0 1 2 3 4 5 6 7

P(k)

k

Maximum for k=5

Bernoulli distribution

Bernoulli distribution

Expected value

Variance

np X

E ( )   

) 1

( )

( X

np p

V    

Bernoulli distribution

Errors in transmission

Example 6.2

Digital channel of information transfer is prone to errors in single bits. Assume that the probability of single bit error is p=0.1

Consecutive errors in transmissions are independent. Let X denote the random variable, of values equal to the number of bits in error, in a sequence of 4 bits.

E - bit error, O - no error

OEOE corresponds to X=2; for EEOO - X=2 (order does not matter)

Example 6.2 cd

For X=2 we get the following results:

{EEOO, EOEO, EOOE, OEEO, OEOE, OOEE}

What is a probability of P(X=2), i.e., two bits will be sent with error?

Events are independent, thus

P(EEOO)=P(E)P(E)P(O)P(O)=(0.1)² (0.9)² = 0.0081

Events are mutually exhaustive and have the same probability, hence

P(X=2)=6 P(EEOO)= 6 (0.1)² (0.9)² = 6 (0.0081)=0.0486

Errors in transmission

Example 6.2 continued

Therefore, P(X=2)=6 (0.1)² (0.9)² is given by Bernoulli distribution

! 6 2

! ) 2 (

! 4 2

4    

 





1 . 0 ,

4 , 3 , 2 , 1 , 0 ,

) 1

4 ( )

(   

 



 



 

 p p

x p

x x X

P

P(X = 0) = 0,6561 P(X = 1) = 0,2916 P(X = 2) = 0,0486 P(X = 3) = 0,0036 P(X = 4) = 0,0001

Errors in transmission

Mean:

P(X = 0) = 0,6561 P(X = 1) = 0,2916 P(X = 2) = 0,0486 P(X = 3) = 0,0036 P(X = 4) = 0,0001

Errors in transmission –

calculation of mean and variance

Variance:

Geometric distribution

The height of the line at x is (1-p) times the height at the line at x-1. That is, the probabilities decrease in a geometric progression. The distribution acquires its name from this result.

Geometric distribution

Lack of memory property (the system will not wear out): A geometric random variable has been defined as the number of trials until the first success. However, because the trials are independent, the count of the number of trials until the next success can be started at any trial without changing the probability distribution of the random variable.

Example 6.3 In the transmission of bits, if 100 bits are transmitted, the probability that the first error, after bit 100, occurs on bit 106 is the probabability that the next six outcomes are OOOOOE and can be calculated as

1 . 0 ,

) 1

( )

6

( X   p

 p

p  P

This result is identical to the probability that the initial error occurs on bit 6.

Poisson’s distribution

We introduce a parameter λ=pn (E(X) = λ)

x n x

x n x

n n

x p n

x p x n

X P









 

  

 

 



 



 



 



 



 



 

  

1 )

1 ( )

(

Let us assume that n increases while p decreases, but λ=pn remains constant. Bernoulli distribution changes to Poisson’s distribution.

1 ! )

( lim

lim ^{P X x n} _{x n n} ^e _x

x n

x

 



_

 

 

  

 

 



 

 

 



Consider the transmission of n bits over a digital communication channel. Let the random variable X equal the number of bits in error. When the probability that a bit is in error p is constant and the transmissions are independent, X has binomial distribution.

It is one of the rare cases where expected value equals to variance:





 np X

E ( )

Why?



    



np np np X

V ( )

lim

(

)

0 , 2

Poisson’s distribution

Poisson’s distribution

Poisson’s distribution

Poisson’s distribution

0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 0,4

0 5 10 15 20 25

lambda=1 lambda=5 lambda=10

x

p(X)

Bernoulli n=50; p=0.02

Poisson:

λ=1 0

1 2 3 4 5 6

0.364 0.372 0.186 0.061 0.014 0.003 0.000

0.368 0.368 0.184 0.061 0.015 0.003 0.001

(x- integer, infinite; x 0) For big n Bernoulli distribution resembles Poisson’s distribution

Poisson’s distribution